You know how when you adjust a mixing board, pushing one slider up can mask the fact that another slider is quietly creeping in the opposite direction? That's the core mechanism here. U.S. consumption inequality declined overall between 2018 and 2022 — but this paper pulls apart the individual faders to show that different forces were pushing in opposite directions simultaneously, and the net decline hides a wealth-driven wedge that got worse. The committed claim: using distribution regression on the Consumer Expenditure Interview Survey, Anand and Liu decompose changes in U.S. consumption inequality into components attributable to individual variables, constructing counterfactual distributions where specific factors are held constant between 2018 and 2022. This is an application of established distributional decomposition methods to a new empirical question — not a methodological first, but a clean identification of which forces moved the needle during a historically unusual period. The key finding splits into three layers. First, changes in the conditional distribution of consumption — how much people at each point in the distribution spend given their characteristics — explain most of the observed decline in consumption inequality among male-headed households. Second, rising asset holdings significantly increased consumption inequality across all measures. Third, changes in a set of household characteristics (demographics, employment composition) significantly reduced inequality. The assets finding is the headline: pandemic-era asset inflation (stocks, housing) didn't just widen wealth gaps — it widened spending gaps too. The method belongs to the distribution regression family, specifically the counterfactual decomposition approach where you estimate the full conditional distribution of consumption, then swap in pre-pandemic coefficients or covariate distributions to isolate each factor's contribution. This is the Chernozhukov-Fernández-Val-Melly lineage of distributional methods, not a simple Oaxaca-Blinder at the mean. The data source — the CE Interview Survey — is the workhorse U.S. consumption dataset, and the authors explicitly focus on well-measured consumption components robust to measurement error, which is a real concern in consumption data where imputation and recall bias are endemic. On integrity: the validation is empirical decomposition on survey microdata, not experimental. There's no pre-registration, no randomization — this is observational economics. The strength is the public data source (CE Survey is accessible) and the focus on measurement-robust components. The weakness is that counterfactual decompositions always carry the assumption that you can meaningfully hold one factor constant while changing another, which in a pandemic where everything moved simultaneously is a real stretch. The choice to focus on male-headed households narrows the scope considerably and raises questions about generalizability. The obvious next experiment: female-headed households, or better yet, the full sample with household-head gender as an interaction term rather than a sample split. The paper restricts to male-headed households without a fully satisfying explanation for why the decomposition isn't run on the full population. My honest read: the decomposition may not have produced clean results for other subsamples, or the sample sizes for female-headed households in the CE Survey may have been too thin for stable distributional estimates. This is likely (a) — data limitations — combined with a bit of (b) — the full-sample story may have been muddier. For the field, this paper is a solid empirical contribution within a well-established methodological tradition. It won't change how anyone does decomposition, but it provides useful evidence for the policy debate about whether pandemic-era fiscal and monetary interventions compressed or widened inequality in the dimension that matters most — actual spending. The asset-holdings finding is the piece worth tracking: if asset-driven consumption inequality persists as pandemic-era wealth effects normalize, that's a structural shift, not a temporary blip.
